Fifty years ago orthopedic surgeons didn’t have to worry about the cost of their treatments…there likely was no one looking over their shoulder. Now, however, with increased scrutiny of both the cost and quality associated with orthopedic procedures, the field is paying more attention to the bottom line…and getting more than a little help from administrative data sets.
“Clinicians and clinician scientists have a responsibility to determine not only the quality, but the cost effectiveness of any given procedure, ” says Dr. Kevin Bozic, an orthopedist at the University of California, San Francisco (UCSF) and an acknowledged health policy expert. “By utilizing administrative claims data sets—information submitted as part of an insurance claim as a result of a patient encounter—we can link clinical information with financial information. This means that we can examine specific diagnoses and procedures, along with their associated costs.”
Rich Sets of Data
“Administrative claims data is a rich source of clinical and financial information, ” states Dr. Bozic, “including information about multiple diagnoses, multiple encounters and procedures. As our healthcare system gets more complex, the data required for payment is more sophisticated—good news for researchers because we can access that data and use it to study healthcare delivery processes and outcomes. For instance, in 2005, an optional modifier code for total hip replacement (THR) procedures was added which allows surgeons and hospitals to specify which type of hip replacement bearing surface was used in the procedure. Having this information from the administrative claims allows us to study trends in utilization and outcomes associated with specific THR bearings.”
If you’re going to make bold statements about what works and what doesn’t, ideally you would have backup from data…lots of data. Dr. Bozic: “The primary advantage of these data sets is that they capture the entire US population. You can, for example, look at the entire Medicare population or single out those who are privately insured. The sample sizes are enormous, therefore the generalizability of the findings can be very broad. This is in contrast to randomized, controlled trials (RCTs) which tend to be focused on a much smaller population and aim to answer a specific clinical question. Although RCTs provide the highest level of evidence regarding the efficacy of a particular treatment, the generalizability of the results can be limited by the small numbers of patients who are studied in an idealized setting—which may or may not be like real life.”
It’s the scientific version of yin versus yang. “This gets to the questions of efficiency and effectiveness, ” states Dr. Bozic. “How an intervention performs in real life is the fundamental issue, such that even if something works in 100 patients who are observed closely, they may or may not be treated in the same way in every practice in the U.S.
The broad generalizability of the administrative data sets comes in part from the fact that they capture a large cross section of patients in different settings, as opposed to randomized controlled studies which are typically conducted in academic practice settings. The former include rural and urban settings, teaching and nonteaching hospitals, different states, etc.
Working Through Limitations
With all of their flexibility and richness, administrative data sets have their limitations. “The disadvantage of this data is that until 2005, the codes used to represent different types of hip replacements and failure of hip replacements were vague. There was only one code used for all causes of hip replacement failures and all types of revision hip replacements. My colleagues and I approached the Centers for Medicare and Medicaid Services, and were able to convince them to add more administrative codes such that we now have a series of codes related to hip and knee replacement procedures which specify the cause of failure and type of revision procedure performed.”
But will the codes—and other information—arrive on the page correctly? “This information is not typically submitted, vetted, or reviewed by a clinician. It is captured by those on the administrative end of things, and the accuracy is therefore at times in question. Additionally, it may or may not accurately reflect the clinical record in that every single clinical condition does not have a corresponding administrative code; the relevance of the code may not be sufficiently nuanced. All of this detailed data is only getting more complex, and many clinicians are not familiar with what it means, how to use it, or how to access it.”
Learning this process, says Dr. Bozic, could involve an informal apprenticeship, as well as hiring people outright.
Those interested in pursuing this line of research should join up with methodologists who have the related expertise. You can purchase all of the data yourself, but this gets costly. If you want 10 years’ worth of data, at $10, 000 per data set, that’s $100, 000. It’s more efficient to find a group of methodologists who own and have experience with the data. Our research group includes both clinicians, who develop clinically appropriate research questions and hypotheses, and methodologists, who understand how to mine and analyze the data.
Delineating a difference between two kinds of data, Dr. Bozic notes, “Clinical data and administrative data are complementary, in that clinical data can often be used to evaluate the efficacy of a procedure (e.g., how a procedure performs under controlled circumstances), while administrative data can be used to evaluate the effectiveness of a procedure (e.g., how a procedure performs under real world conditions)”.
As far the accuracy of the data sources, Dr. Bozic states, “Clinical data that comes from a chart review is not necessarily more or less accurate than administrative data. Someone has abstracted that data, put it in a spreadsheet, and coded it in terms of how many people had infections, etc. Because of concerns about fraud and abuse, administrative data is strictly audited by several parties, including payers.”
Data Mining and its Future Value
In taking this “clean” data on the road, Dr. Bozic has noticed a new trend in the research world. “For the last several years, I have presented studies at AAOS in which I’ve utilized administrative data sets. As the Program Chair for the American Association of Hip and Knee Surgeons, I have seen that approximately 25% of the abstracts selected for presentation at the 2009 Annual Meeting came from administrative data, a significant increase from, say, five years ago. There has obviously been a shift toward understanding and acceptance of the value of this type of data. The reality is that you need both types of information…administrative data and data from well controlled clinical studies.”
Next, he would like to take it to the community—the community practitioners, that is. “Going forward I think we will see more robust data sets, accompanied by an increased interest in this type of research tool. And ideally, rather than spending tens of thousands of dollars on the data, those with less training will be able to access the data in a less expensive, more usable, form. A significant challenge that remains to be solved is to be able to track patients over time in administrative data sets using a unique identifier, something which involves issues of patient privacy.”
You could say that a good clinician scientist wears a miner’s hat…he or she must extract all of the information possible from a given set of data. Dr. Bozic:
There are still a multitude of questions that can be asked using administrative data. For example, what are the risk factors for infection following orthopedic procedures, and which types of treatment are most likely to be successful? The ultimate advantage of this data is that you can connect clinical and financial information, meaning that we can look at types of diagnoses and procedures along with their associated costs. That, ideally, will help drive down costs and make orthopedic care more accessible to a greater number of people.
